Perturbation of the Tumor Necrosis Factor–Related Apoptosis-Inducing Ligand Cascade in Ovarian Cancer: Overexpression of FLIP_L and Deregulation of the Functional Receptors DR4 and DR5

Despite great clinical and research efforts, the etiology of ovarian cancer remains poorly understood. A majority (90%) of these tumors stem from ovarian surface epithelium, which is similar to the mesothelial lining of other abdominal organs. Beyond family history, the most important risk factors for ovarian cancer include early menarche, late menopause, nulliparity, and in contrast, the use of anovulatory drugs that reduces the risk. The epidemiologic observation that all factors that limit the number of ovulations over lifetime are protective against ovarian cancer resulted in the so-called incessant ovulation theory decades ago (1).

Over the past several years, the molecular mechanisms behind the genetic predisposition for ovarian cancer have been elucidated. Germ line mutations in BRCA1 and BRCA2, the genes responsible for hereditary breast and ovarian cancer and in hMLH1 and hMSH2, responsible for hereditary nonpolyposis colon cancer, are also the basis for a majority of cases of familial ovarian cancer. Originally identified by positional cloning, all these genes were later shown to be involved in the maintenance of genome integrity, implying that DNA repair mechanisms are under considerable challenge in ovarian tissue. Induction of apoptosis plays a pivotal role in cellular homeostasis, and in this scenario, the impairment of any of its components may result in accumulation of genetic defects, preceding cancer development (2).

Apoptosis is induced either by binding of ligands to specific death receptors on cell surface or by unspecific cellular stress, which promotes cytochrome c release from mitochondria and the formation of the apoptosome. These pathways converge at the level of the effector caspases that bring about cell death by cleavage of various cellular substrates (3). Tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) was identified as an unique death ligand in respect to its ubiquitous receptor expression and the lack of cytotoxic effects in normal tissue (4, 5). TRAIL induces apoptosis by binding to its two death-inducing receptors, death receptor 4 (DR4, TRAIL-R1) and DR5 (TRAIL-R2, KILLER; ref. 6), recruiting Fas-associated death domain and procaspase-8 (7–9) to their intracellular death domains, hereby forming the death-inducing signaling complex and initiating apoptosis. Regulation of TRAIL-induced apoptosis was further attributed to cellular FLICE inhibitory protein (FLIP), which can efficiently block caspase-8 cleavage (10) due to a structural similarity to caspase-8. Two forms of FLIP have been identified to date, both containing two death effector domains, enabling them to interact with Fas-associated death domain and/or caspase-8. The longer isoform, FLIP_L, also contains a rudimentary caspase-like domain without catalytic activity. High expression of FLIP has been correlated with TRAIL resistance in various tumor types (11, 12), including ovarian cancer (13–15). The postulated physiologic role of FLIP has been in immune escape (16, 17), hereby conferring an advantage for tumor cells. Moreover, a gene expression profiling study identified FLIP as one of the genes differentially expressed in ovarian cancer compared with normal ovarian epithelium (18). In addition, two membrane-bound decoy receptors unable to activate the apoptotic cascade, DcR1 (TRAIL-R3, TRID) and DcR2 (TRAIL-R2, TRUNDD; ref. 6) exist. Missing (DcR1) or truncated (DcR2) death domain and consequent inability of forming the death-inducing signaling complex hinders apoptotic activity of these receptors upon TRAIL binding. Interestingly, genes coding for all membrane bound TRAIL receptors are located on the short arm of chromosome 8 (8p21.3 according to the University of California, Santa Cruz Genome Browser Database, http://genome.ucsc.edu/), a region often found affected by genomic aberrations in ovarian and other cancers. Overexpression of decoy receptors was related to in vitro TRAIL resistance, although they do not seem to determine TRAIL resistance or sensitivity under physiologic conditions (19).

Facing the possibilities of TRAIL therapy in ovarian cancer, we are in search for determinants of TRAIL resistance in ovarian cancer (14). We previously showed that DR4 is functionally silenced in a considerable number of ovarian cancer patients (20). In this study, we determined the protein expression levels of TRAIL, death-inducing TRAIL receptors DR4 and DR5, and FLIP_L in ovarian cancer epithelium of 68 ovarian cancer patients and assessed their effect on overall survival.

Ovarian tissue microarray. Ovarian tumor biopsies were obtained from patients undergoing initial staging or debulking laparotomy, and before freezing, sample mass was reduced by dissecting normal tissue. For immunohistochemical studies, paraffin material available from primary diagnosis was used. Patients gave informed consent according to the criteria used at the Medical University of Vienna. Relevant clinical information was gathered, and tissue samples and clinical data were anonymized. A tissue array was assembled by taking core needle “biopsies” from specific locations in the preexisting paraffin-embedded tissue blocks and reembedding them in an arrayed “master” block, using techniques and an apparatus developed by Beecher Instruments, Inc., Micro-Array Technology (Sun Prairie, WI). To achieve good representation of the tumor, three biopsies of tumor material and one of nonmalignant, regular, mostly stromal ovarian tissue were selected from each patient sample. Using this technology, each tissue sample is treated in an identical manner, and the entire cohort is analyzed in one batch on a single slide. Reagent conditions are identical for each case, as are incubation times and temperatures, wash conditions, and antigen retrieval if necessary. A 4- to 5-μm paraffin section of the tissue microarray was deparaffinized and rehydrated, and subsequently, the section was treated with 0.2% H₂O₂/PBS (pH 7.4) to quench endogenous peroxidase activity. After blocking with 2% normal goat serum for 30 minutes, the section was incubated at 4°C overnight with primary antibody [rabbit polyclonal DR4 (H130), goat polyclonal DR5 (C-20), TRAIL (K-18), and FLIP_L (C-19) antibodies; Santa Cruz Biotechnology, Inc., Santa Cruz, CA] in 1% normal goat serum. For TRAIL, DR5 and FLIP_L staining normal rabbit serum was used. Horseradish peroxidase–linked anti-rabbit (NA 934, Amersham Biosciences, Buckinghamshire, United Kingdom) secondary antibody and anti-goat (A4174, Sigma-Aldrich, St. Louis, MO) secondary antibody, respectively, was applied for 60 minutes followed by 2 to 4 minutes of incubation in diaminobenzidine substrate (DAKO Liquid DAB+, DakoCytomation, Glostrup, Denmark). The section was counterstained in hematoxylin for 90 seconds and mounted under a coverslip. Consecutively, tissue staining was assessed by cell counting and categorized into three groups according to the percentage of positive cells (<10%, 10-30%, and >30% stained cells). Intensity of the staining was rated on scale of 1 to 3, with 1 for weak staining, 2 for moderate, and 3 for strong staining. Immunohistochemical analysis revealed a combined membrane and cytoplasmatic staining pattern for DR4 and DR5 and a mainly cytoplasmatic one for TRAIL and FLIP_L. All histologic and morphologic data included in Table 1 and all additional immunohistochemical evaluations described below were verified on the tissue array by two individual pathologists.

Statistical methods. For statistical analyses, the immunohistochemical stainings were considered as binary variables (positive versus negative). In the case of TRAIL and FLIP_L, moderate and strong staining in >10% of tumor cells was considered positive, in the case of DR4 and DR5, only moderate and strong staining in >30% was counted as positive. Furthermore, the following clinical variables were used as binary variables for statistical analysis: histology (serous versus nonserous tumors), stage [Federation Internationale des Gynecologistes et Obstetristes (FIGO) I/II versus III/IV], grading (grade 1 versus 2/3), and age at diagnosis (<50 versus ≥50).

To analyze the dependence of the immunohistochemical variables and their association with clinical variables, the χ² test was used. In case of very low cell frequencies (expected frequency < 5), the Fisher's exact test was used. Survival was estimated using the Kaplan-Meier method, and Cox regression analyses were done to evaluate the influence of stage, grade, and the immunohistochemical variables on survival. Univariate regression models were estimated for stage, grade, and each of the immunohistochemical variables. The stepwise method was used to add only the statistically significant variables to the multiple regression model, including the variables stage and grade. Based on previous observations, additional survival analyses according to Kaplan-Meier were done in the subgroup of advanced-stage (FIGO III/IV) patients only (21).

Ps < 0.05 were considered statistically significant.

The patient cohort investigated comprised 68 women diagnosed with epithelial ovarian cancer and treated in a single institution by debulking laparotomy followed by a standard chemotherapeutic regimen, consisting of carboplatin/paclitaxel (39 patients) cisplatin/cyclophosphamide (12 patients), other protocol (10 patients), or no further therapy in seven patients (FIGO Ia-Ib, G₁). Clinical data at the time of diagnosis as well as the clinical outcome were available for this analysis. Patient characteristics, including FIGO stage, tumor grading, and age, are depicted in Table 1 and are representative for an unselected ovarian cancer population. The median time of clinical follow up was 63 months (range, 9-169 months).

Immunohistochemical analysis of tumor necrosis factor–related apoptosis-inducing ligand, death receptors 4 and 5, and FLICE inhibitory protein. TRAIL was expressed by epithelial ovarian cancer cells in 40.4% at varying levels. In stromal cells, TRAIL staining was present in 43.9%. In both cases, the staining was cytoplasmatic. TRAIL staining in the microenvironment of the tumor was slightly more common than in tumor cells and independent (χ² test, P = 0.620; Fig. 1) from each other. We observed a statistically significant relationship between epithelial TRAIL expression and FLIP_L (P < 0.0001), although no correlation of stromal TRAIL and FLIP_L could be observed. Furthermore, we found a statistically significant higher DR5, but not DR4, expression in ovarian cancer cell epithelium, which also displayed elevated TRAIL expression (P = 0.041). However, a statistically significant coexpression of DR4 and DR5 in ovarian cancer cells (P = 0.021) was also the case. Frequencies of all immunohistochemical stainings are summarized in Table 1 and examples are shown in Fig. 2A-F.

FLICE inhibitory protein expression is nonredundant with reduced death receptor 4 and/or 5 expression. Tumor cells could escape TRAIL-induced apoptosis either by down-regulation of functional receptors or overexpression of inhibitory proteins. To get at least an indirect evidence whether these mechanisms would be relevant in ovarian cancer, patients were classified according to the expression levels of DR4 and DR5 into groups of regular receptor expression (strong staining on both receptors or moderate staining on one and strong staining on the other receptor in >10% of tumor cells) and diminished expression (all other combinations) and related to FLIP_L expression. Direct comparison of DR4/DR5 and FLIP_L showed that only 4 of 31 (12.9%) tumors, which showed FLIP_L overexpression, also had diminished TRAIL receptor expression compared with more than one third (14 of 37, 37.8%) of tumors with low or missing FLIP_L and reduced DR4 and DR5 expression levels (Fig. 2G). This difference was statistically significant (P = 0.020). Taken together, not less than 66.2% of the patients had some alteration in the TRAIL pathway, displaying either reduced DR4/DR5 expression (20.6%), increased FLIP_L expression (39.7%), or both (5.9%) in the population studied.

Analysis of tumor necrosis factor–related apoptosis-inducing ligand, death receptors 4 and 5, and FLICE inhibitory protein and relevant clinical variables. As a next step, the correlation of the expression profiles of the individual genes to clinical variables of prognostic value, like histologic subtype, stage, grade, and age, were investigated. Interestingly, higher epithelial DR4 expression levels correlated strongly with tumor grading (P < 0.0001). As previously shown, we observed a high percentage of reduced epithelial expression of DR4 in patients younger than 50 years of age at diagnosis (P = 0.05; ref. 20). There was a strong tendency towards epithelial FLIP_L expression in early-stage, nonserous tumors (P = 0.024 and P = 0.013).

Survival analysis in ovarian cancer patients. We aimed to determine the effect of TRAIL and its receptors on survival of ovarian cancer patients. In the Kaplan-Meier survival analysis of 54 patients, we found a statistically significant (log-rank test) favorable overall survival for low grade (grade 1 versus grade 2-3; P = 0.047) and low stage (FIGO I/II versus III/IV; P = 0.0008). In univariate regression analysis, neither epithelial DR4 and DR5 expression nor TRAIL or FLIP_L levels in ovarian cancer epithelium and/or stroma were statistically significant predictors for survival benefit in our study population. Stepwise multivariate Cox regression analysis revealed only stage (FIGO I/II versus III/IV; P = 0.013) as a statistically significant independent factor for survival outcome in our cohort of ovarian cancer patients.

Stromal tumor necrosis factor–related apoptosis-inducing ligand expression and ovarian cancer survival. TRAIL was previously shown to confer a survival benefit when determined on the mRNA level in advanced ovarian cancer (21). We observed a survival benefit of high TRAIL expression in advanced-stage patients with markedly elevated stromal expression (P = 0.049, log-rank test; Fig. 3). In contrast, epithelial expression of TRAIL, which was completely independent from stromal expression (P = 0.620), did not have an effect upon overall survival (P = 0.869; Fig. 3).

We investigated the expression of TRAIL and its two apoptosis-inducing receptors DR4 and DR5 in a representative ovarian cancer population by immunohistochemistry. Using an ovarian cancer tissue microarray, which included for each patient three biopsies of tumor tissue and one of mainly stromal tissue on one single microscopic slide, we were able to get additional insight in the contribution of the TRAIL and its functional receptors to the pathogenesis of ovarian cancer. Furthermore, we assessed in a large-scale analysis the protein expression of the caspase-8-inhibiting protein FLIP_L in ovarian cancer.

Interestingly, we could observe for the first time a significant enhancement of stromal TRAIL expression in the microenvironment of ovarian cancer cells, and this expression was independent from the autocrine expression in cancer cells. It was shown previously that in normal ovarian tissue, stromal TRAIL expression is completely missing but to some extent present in epithelial cells (22). This suggests, that the stromal expression detected by us is most likely a reaction of the microenvironment to the tumor cells. There was no evidence for lymphatic invasion in the stromal tissue, making the stromal cells the most likely source of TRAIL expression. In advanced stages, only the stromal TRAIL expression was a strong predictor of overall survival but not the autocrine expression of TRAIL in the tumor cells. Whereas TRAIL expression was connected previously to overall survival (21), it was assessed by real-time reverse transcription-PCR from mRNA isolated from nonmicrodissected ovarian cancer samples comprising tumor and stromal tissue. By using immunohistochemistry rather than real-time reverse transcription-PCR, we could not only confirm the survival benefit of high TRAIL expression in advanced stage ovarian cancer but also determine the origin of its expression. Only stromal TRAIL expression predicted a significantly better ovarian cancer survival and did not correlate with epithelial expression levels of TRAIL, TRAIL receptors, FLIP_L, nor any histopathologic variables.

It is intriguing that the observed and, in many cases, very strong autocrine expression of TRAIL did not show any influence on the prognosis. This raises the question whether tumor cells can protect themselves from TRAIL-induced apoptosis. We found a markedly elevated FLIP_L expression in a large number (46.3%) of ovarian cancer tissues, a result in accordance with previous observations (23). This expression correlates very closely to the epithelial TRAIL expression, which strongly suggests an autoprotective mechanism of ovarian cancer to escape TRAIL-induced apoptosis. We have shown previously that epigenetic silencing of DR4 by methylation contributes to ovarian cancer and TRAIL resistance (20). Consequently, we observed a loss of expression of DR4 and/or DR5 in a corresponding number of cases, a factor that may contribute to the tolerance of tumor cells against TRAIL. This loss of DR4 and/or DR5 was significantly more frequent in tumors not overexpressing FLIP_L, herewith implying two independent resistance mechanisms in the TRAIL-induced apoptotic pathway occurring in ovarian cancer. Moreover, elevated FLIP_L expression was mostly present in early-stage ovarian cancers and significantly reduced in tumors of advanced-stage patients. This fact could also contribute to the survival benefit of advanced-stage patients with elevated stromal TRAIL expression because of the attenuation of an important apoptotic inhibitor.

Animal studies targeting the clinical feasibility of TRAIL have confirmed the cancer therapeutic potential of TRAIL (24–26) despite previous reports on TRAIL hepatotoxicity (27). Differential activation of TRAIL receptors DR4 and DR5 by soluble and membrane-bound TRAIL has been addressed in some publications (28–30), initiating the evaluation of more specific death receptor ligands (31, 32). In fact, agonistic antibodies to DR4 and DR5 have already entered phase I trials (33, 34), paving the way for novel cancer therapeutic strategies based on TRAIL. These observations point out the possible therapeutic effect of exogenous TRAIL in cancer therapy, particularly in combination with cytotoxic drugs.

Taken together, our data show that in addition to DR4/DR5 down-regulation, TRAIL sensitivity may be significantly altered by FLIP_L overexpression and that elevated stromal TRAIL expression confers a survival benefit for advanced-stage ovarian cancer patients. Currently, standard therapy of advanced ovarian cancer consists of radical surgery combined with adjuvant platinum-based chemotherapy, although the development of drug resistance and disease recurrences remains a major problem (35). The biological effects of TRAIL in ovarian cancer models, its enhancement of chemosensitivity towards standard chemotherapeutic agents and the effect of endogenous TRAIL levels on survival make TRAIL an exciting and auspicious candidate for ovarian cancer therapy.